CN102024146A - Method for extracting foreground in piggery monitoring video - Google Patents
Method for extracting foreground in piggery monitoring video Download PDFInfo
- Publication number
- CN102024146A CN102024146A CN 201010577516 CN201010577516A CN102024146A CN 102024146 A CN102024146 A CN 102024146A CN 201010577516 CN201010577516 CN 201010577516 CN 201010577516 A CN201010577516 A CN 201010577516A CN 102024146 A CN102024146 A CN 102024146A
- Authority
- CN
- China
- Prior art keywords
- prospect
- frame
- background
- foreground
- difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method for extracting foreground in piggery monitoring video. The method comprises the steps of firstly acquiring an initial background frame containing no foreground and a background frame sequence containing the foreground, then carrying out adjacent symmetric frame difference and background frame difference, carrying out foreground motion analysis according to the adjacent symmetric frame difference and the background frame difference, acquiring a status code, updating background and fusing the two frame differences by utilizing the status code so as to acquire a shade containing foreground, calculating background Euclidean distance of local texture operators and S (saturation) and H (hue) channel background frame difference of HSV (hue-saturation-value) colour space, combining shade discriminating conditions, and finally extracting a foreground frame according to result of pixel attribute values. The invention is applicable to piggery monitoring of large-scale piggery farms, has adaptability, robustness and higher foreground target segmentation accuracy and creates favorable conditions for subsequent work which is pig video intelligent analysis.
Description
Technical field
The present invention relates to a kind of video monitoring foreground target extracting method, relate in particular to the fixedly extracting method of monitoring camera-shooting video pig of pig house, specifically be applicable to pig farm intelligent video monitoring technical field, belong to the intelligent video target detection technique.
Background technology
It is with the background removal in each frame in the sequence of frames of video that the video monitoring foreground target extracts, the technology that detects foreground target and split.Foreground extracting method is divided into fixing by cam device and moves two big classes, divides rigid body and non-rigid body two big classes by the foreground target characteristic.Foreground extracting method belongs to the non-rigid body foreground extraction technology of fixing shooting in the pig house monitor video.
Doctor Zhang Min (referring to: open quick. based on the animal behavior of figure's identification automatically analysis and research with use [D]. Hangzhou Zhejiang University biomedical engineering and instrument institute 2005.) adopt the complete foreground target of coloured image threshold value outline acquisition, this method is simple, calculation cost is little, belong to classical background frame difference method, but it is static constant that this method requires background, and background heals the simple division effect better.And the pig house background of reality is difficult to satisfy this condition, because pig house generally is designed to the ventilation and penetrating light building, fine light is strong, and pig is moved indoor, shadow can occur; Gradual change also can take place in cloudy day light in time; When temporary blocking occurring suddenly, the window of projection sunlight can play the indoor illumination intensity sudden change.
People such as professor Xin (Shao B, Xin H. A real-time computer vision assessment and control of thermal comfort for group-housed pigs[J]. Computers and Electronics in Agriculture. 2008,62 (1): 15-21.) utilize on-the-spot dark-background in laboratory and the big characteristics of white pig contrast, the average that adopts continuous 3 frames is as processed frame, the utilization gray level threshold segmentation goes out preliminary prospect, carrying out morphology and area threshold again handles, obtain accurate prospect, this is owned by France in Flame Image Process dividing method acquisition foreground target, is suitable for and only limits to specific experiment condition.Particularly when prospect was moved, the mean value of continuous 3 frames caused the pseudo-impact point of prospect to increase.Directly adopting Flame Image Process to ignore the time domain specification of frame sequence, is not the common method of frame of video target detection.
General frame of video background removal approach commonly used (referring to: Herrero S, Besc ó s J. Background Subtraction Techniques:Systematic Evaluation and Comparative Analysis[C] //Proceedings of the 11th International Conference on Advanced Concepts for Intelligent Vision Systems.Springer-Verlag. 2009,5807/2009:33-42.) be divided into 3 classes, be basic model, as frame difference method, medium filtering; Parameter model is as single Gauss, mixed Gauss model; Nonparametric model is as histogram method, cuclear density method.And mixed Gauss model and nonparametric model calculated amount are big, the method complexity, and real-time is poor, thereby has limited its popularization in actual applications.And medium filtering and single Gauss model can produce smear when foreground target slowly moves; Adjacent frame difference method is calculated simple, but object content easily produces the cavity, and when pig plants oneself, can be made background by mistake; The background frame difference method must be set up the context update model, and above-mentioned 3 class algorithms all can't be eliminated the shadow of prospect.
Hu Yuanyuan (referring to: Hu Yuanyuan, Wang Rangding. based on the motion shadow removal algorithm [J] of local grain unchangeability. computer utility. 2008,28 (012): 3141-3143.) propose to describe the local grain structure and can distinguish foreground target and shadow well with the LBP operator that strengthens.Yet this method will be ineffective when background image and foreground object have similar texture information, and therefore, this has just limited the application of algorithm.
Summary of the invention
The objective of the invention is to overcome the deficiency of existing pig video foreground target extraction method, foreground extracting method in a kind of robustness is good, effective and real-time the is good pig house monitor video is proposed, can keep accurate frame of video prospect edge and eliminate common methods and easily cause prospect interior void, smear and shade and deposit cash resembling, and can suppress electronic noise and ground is water stain and the excreta vestige changes influence to prospect.
The technical solution used in the present invention is: in the pig house zone camera system and computer control system are set, there are adjacent symmetric frame difference processing block module, background frames difference processing frame module, motion analysis processing block module, prospect Fusion Module, context update module, shadow Detection module and foreground extracting module in computer control system inside.Obtain an initial background frame that does not contain prospect by camera system
With the background frames that contains prospect
Sequence also keeps in computing machine.Adjacent symmetric frame difference processing block module input background frames
Continuous 3 consecutive frames in the sequence are with present frame
With forward and backward frame
,
Carry out difference respectively and differentiated two binary images carried out AND operation obtaining the adjacent symmetric frame difference
Background frames difference processing frame module input present frame
With the initial background frame
, be output as the background frames difference of binaryzation
, select the B passage of RGB color space to carry out the background frames difference
Motion analysis processing block module input background frames difference
With the adjacent symmetric frame difference
Carry out motion state and differentiate, output present frame prospect state encoding
,
It is present frame
Carry out the motion state of adjacent symmetric frame difference computing,
Be present frame
Frame carries out the scene state of background frames calculus of differences; The ratio that accounts for total image area when moving region area foreground moving less than 0.04~0.12 time is slow or static or do not have a prospect; When ratio that the prospect area accounts for total image area background illumination greater than 0.6~0.8 time takes place to become suddenly, do not have prospect in the scene less than 0.03~0.07 the time; The prospect state encoding of prospect proper motion
Be 00, the slow prospect state encoding of motion of prospect
Be 10, do not have the prospect state encoding of prospect
Be 11, the prospect state encoding of background illumination sudden change
Be 02.The prospect Fusion Module is with the background frames difference
With the adjacent symmetric frame difference
Carry out inclusive-OR operation, obtain containing the prospect binary map of shade through filtering, morphological operation and connection marking operation
Context update module input present frame prospect state encoding
, the initial background frame
, preceding symmetrical frame
With its prospect binary map
, present frame
With its prospect binary map
, output is background frames
When the prospect state encoding
Be to rebuild brand-new background frames at 11 o'clock
When the prospect state encoding
Be 00 or 10 o'clock, with present frame
Remove the prospect binary map
Back remaining area pixel replaces the initial background frame
Respective pixel is by the prospect binary map
The background area of blocking is the initial background frame
The respective pixel value compensates the prospect binary map again
Present frame
Prospect edge pixel and corresponding preceding frame
The average of margin of image element; When the prospect state encoding
It is 02 o'clock, by preceding frame
Remove the prospect binary map
The present frame of back remaining area correspondence
Pixel replaces the initial background frame
Corresponding background pixel, the initial background frame
Middle prospect binary map
The background area of blocking is the initial background frame
The respective pixel value compensates the prospect binary map again
The preceding frame of prospect edge correspondence
Pixel and corresponding present frame
The average of margin of image element.The shadow Detection module will contain the prospect binary map of shade
Corresponding present frame
Pixel value is got
RGBThe value of color model B passage adopts 8 neighborhood territory pixels to make up the local grain zone, calculates the local grain construction operator
And compare two construction operators
The Euclidean distance of value, if less than 0.11, then this pixel is the direct-shadow image vegetarian refreshments, otherwise is the foreground pixel point; The prospect binary map that will contain the direct-shadow image vegetarian refreshments again
Corresponding present frame
Pixel value
Shade is differentiated the saturation degree of prospect shadow region in the hsv color space
The background area of not upgrading with correspondence
Difference less than 0.22 and the colourity of shadow region
The background area of Geng Xining not
Difference then is judged as the direct-shadow image vegetarian refreshments less than 0.08, otherwise is the foreground pixel point.Computing machine by foreground extracting module with present frame
Middle pixel value
Be that 0 zone is left present frame
Original pixel value, rest of pixels is shown as black background, and the foreground extraction that then will not contain shade is come out.
The invention has the beneficial effects as follows:
1, the background frames difference with adjacent symmetric frame difference and adaptive background renewal merges, and by the shadow Detection algorithm, can not only accurately detect the sport foreground target area again, is applicable to that also slow the or static foreground target of motion detects.And shadow Detection algorithm simultaneous texture structure
LTOperator and hsv color space shade double check strengthen the recognition capability to shade.
2, when carrying out the frame processing, all computing modules all are made of the calculated performance simple algorithm, and algorithm is carried out flow process and adopted potential concurrent structure, concurrent execution when being easy to algorithm and being transplanted on the multiprocessor hardware platform realizes that reliable real-time video two field picture handles.
3, the present invention is applicable to the pig house monitoring on scale pig farm, in pig house indoor no matter single goal or multiple goal, no matter background light is soft gradual change or strong sudden change, no matter pig is the station or crouches or stop-go, can both eliminate cavity, smear and shade, and can suppress electronic noise and ground is water stain and the excreta vestige changes influence to prospect, obtain accurate foreground target, have adaptivity, robustness and higher foreground target segmentation precision, for the analysis of follow-up work pig video intelligent creates favorable conditions.
Description of drawings
Below in conjunction with the drawings and specific embodiments the present invention is described in further detail;
Fig. 1 is a foreground extracting method FB(flow block) in the pig house monitor video of the present invention;
Fig. 2 is the process flow diagram of shadow Detection algorithm among Fig. 1.
Embodiment
The present invention is when implementing, in scale hoggery pigsties zone camera system and computer control system are installed, camera system is linked to each other with computer control system, and computer control system inside has adjacent symmetric frame difference processing block module, background frames difference processing frame module, motion analysis processing block module, prospect Fusion Module, context update module, shadow Detection module and foreground extracting module.The ground in scale hoggery pigsties zone is concrete floor, leaves the part excreta above, and whole video background light is brighter.Carry out adjacent symmetric frame difference and background frames difference with parallel algorithm, output variable according to adjacent symmetric frame difference and background frames difference, carry out the foreground moving analysis, obtain state encoding, being that controlled condition is parallel with the status code carries out that adaptive background upgrades and preceding two frame differences merge acquisition and comprise shade in interior prospect, the shadow Detection module is calculated local grain operator (Local Texture with parallel algorithm
LT) Euclidean distance and expression form and aspect, saturation degree and brightness
HSVColor space
S,
HPassage prospect frame is poor, and simultaneous shade criterion is last, according to result's sign of pixel property, eliminates shade, extracts the present frame that only contains prospect.Concrete steps are as follows, referring to Fig. 1:
1, obtains the initial background frame
Computing machine obtains a barnyard scape frame of video that does not have pig, i.e. an initial background frame that does not contain prospect by fixing shooting
, and with this initial background frame
Keep.Afterwards, have a pig slowly to come into this zone, stopping bows gnaws ground a little while, and of short duration defecation phenomenon is arranged, and leaves this zone then; In this time period of pig activity, obtain to contain the background frames of prospect by fixing camera system
Sequence.Background frames pig
In the sequence, a frame that does not contain pig is arranged, also this can not contained the frame of pig as the initial background frame
2, adjacent symmetric frame difference
Background frames pig
In the sequence, computing machine is imported continuous 3 consecutive frames by adjacent symmetric frame difference processing block module
,
,
, output is middle present frame
The moving region of binaryzation.In continuous 3 two field pictures of frame sequence, in order to extract present frame
The motion target edges, the present invention adopts the adjacent symmetric frame difference
, promptly present frame and front and back adjacent symmetric frame carry out difference respectively, and differentiated two binary images are carried out AND operation, obtain the result and are present frame moving target edge, i.e. adjacent symmetric frame difference
See formula (1).The adjacent symmetric frame difference
Be promising 1 some region, belong to the foreground moving zone.The threshold value of binaryzation
Adopt maximum variance between clusters that Japanese Otsu proposes (referring to Otsu N. A threshold selecti on method from gray-level histograms[J]. IEEE Transactions on System Man and Cybernetics. 1979,9 (1): 62-69.), also need compensate a constant relevant with the camera system noise
, generally between 0.06~0.3, value of the present invention is 0.1 to its value, filters noise between frames when guaranteeing not contain foreground target.This algorithm has strengthened the moving target margin signal, widens the difference of target and background residual noise, eliminates the background texture that moving target blocks or reappears.
3, background frames difference
Background frames difference processing frame module in the computer system is input as present frame
And initial background frame
, be output as the background frames difference of binaryzation
Background frames difference block of the present invention is in order to increase the internal signal of prospect.Color of prospect pig and texture and background difference are bigger, and segmentation effect is obvious.Select
RGB(Red Green Blue,
RGB) color space
BPassage carries out present frame
With the background frames difference
, the background frames difference
The threshold value of binaryzation
Adopt maximum variance between clusters, also need compensate a constant relevant with the camera system noise
, generally between 0.06~0.20, value of the present invention is 0.12 to its value, filters noise between frames when guaranteeing not contain foreground target, sees shown in the formula (2).If
Being 1, then is foreground area, otherwise, be background.
4, motion analysis
Motion analysis processing block module in the computer system is input as the background frames difference
With the adjacent symmetric frame difference
, be output as present frame prospect state encoding
, can carry out motion state differentiation and output present frame
The processing controls condition.Because adjacent symmetric frame difference
Can only obtain to comprise the moving region of shade, area when the moving region in interior prospect
Account for total image area
N(
I) ratio
Less than certain less threshold value
e 1 The time, it is relevant with the moving target number with frame per second, and generally 0.04~0.12, the present invention gets 0.07, thinks that then foreground moving is slow or static or does not have prospect.Because background frames difference
Can obtain to comprise shade, when the prospect area in interior foreground area
Account for total image area
N(
I) ratio
Greater than certain big threshold value
e 2 The time, generally 0.6~0.8, the present invention gets 0.7, shows that this frame does not have foreground target or illumination to take place to change suddenly, only carries out the context update operation, enters next frame.
Less than less threshold value
e 3The time, it is relevant with the camera system noise, and generally 0.03~0.07, example of the present invention gets 0.03, then thinks not have prospect in the scene.By formula (3), (4), (5), (6), obtain present frame
The prospect state encoding
, can be by the corresponding follow-up controlled condition of table 1 output.
Be present frame
Carry out the motion state that the computing of adjacent symmetric frame difference can be represented,
For
Frame carries out the scene state that the background frames calculus of differences can be represented.Implement different period state encodings
Be followed successively by 11,00,10,00,11.
(4)
(6)
Table
State encoding
S t S t 'And controlled condition output
Condition discrimination and present frame are handled | ||
0 | 0 | The prospect proper motion then carries out context update, and prospect merges, shadow removal |
1 | 0 | Prospect is slowly moved, and then carries out context update, and prospect merges, shadow removal |
1 | 1 | Do not have prospect, only carry out the context update operation, enter next frame |
0 | 2 | The context update operation is only carried out in the background illumination sudden change, enters next frame |
5, prospect merges
Prospect Fusion Module in the computer system carries out the background frames difference
With the adjacent symmetric frame difference
Merge, promptly the two carries out the prospect binary map that inclusive-OR operation obtains containing shade
, adjacent symmetric frame difference like this
The foreground information of losing obtains fine compensation, the prospect binary map
It is the coarse foreground area that contains shade.At last, through filtering elimination isolated point noise, morphological operation and be communicated with marking operation and thoroughly eliminate little cavity in the foreground area, excrement piece that small size threshold value elimination picture has just produced and urine patch obtain containing the accurate prospect binary map of shade
Prospect fusion treatment frame module of the present invention is input as condition control signal, i.e. present frame
The prospect state encoding
, binaryzation the background frames difference
With the adjacent symmetric frame difference
, be output as the prospect binary map that contains shade
, wherein,
Be 11 o'clock, expression does not have pig to have present frame
Do not need to carry out prospect and merge, change next frame and handle; Be be expressed as in 10,00 o'clock pig slowly, proper motion, can not eliminate interior void behind the adjacent symmetric frame difference, and be difficult to find static foreground target, and the background frames difference can't obtain the accurate edge of moving target, solution is to adopt the background frames difference
With the adjacent symmetric frame difference
Merge, promptly the two carries out inclusive-OR operation and obtains coarse background frames difference
With the adjacent symmetric frame difference
The foreground information lost of adjacent symmetric frame difference obtains fine compensation like this, at last, through filtering elimination isolated point noise, morphological operation and connection marking operation are thoroughly eliminated the little cavity in the foreground area, excrement piece that small size threshold value elimination picture has just produced and urine patch obtain prospect binary map accurate, that contain shade
6, context update
Because real background is not static, so must background upgrade, the context update module in the computer system of the present invention is safeguarded a background model by concurrent program all the time.The input of context update processing block module is a present frame
The prospect state encoding
, the initial background frame
, preceding frame
With its prospect binary map
, present frame
With its prospect binary map
, be output as background frames
Initial background is the initial background frame that does not contain prospect in the video sequence leading portion
The prospect state encoding
Be 11 o'clock, expression does not detect prospect, by present frame
All pixel replaces the whole pixels of background, is equivalent to rebuild brand-new background frames
Be 00 or 10 o'clock, expression detects prospect, then exists
The remaining area pixel replaces the initial background frame after removing foreground area
Respective pixel is by the prospect binary map
The background area of blocking is the initial background frame
The respective pixel value compensates the prospect binary map again
Present frame
Prospect edge pixel and corresponding preceding frame
The average of margin of image element; If
Be 02 o'clock, show to detect the background illumination sudden change, then by preceding frame
Remove foreground area
The present frame of back remaining area correspondence
Pixel replaces former
Corresponding background pixel, former
Middle prospect
The background area of blocking is
Frame respective pixel value compensates prospect again
Prospect edge correspondence
Pixel and corresponding present frame
The average of margin of image element.
7, shadow Detection:
Shadow Detection module in the computer system of the present invention contains two concurrent shadow Detection sub modular structures, and the purpose of using two submodules is for strengthening the robustness to the shadow Detection algorithm.
g 4 | g 3 | g 2 |
g 5 | g 0 | g 1 |
g 6 | g 7 | g 8 |
The prospect binary map that will contain shade
Corresponding present frame
Pixel value is got
RGB(Red Green Blue,
RGB) value of color model B passage, adopt 8 neighborhood territory pixels with 1 pixel of central pixel point distance to make up local grain and described the zone, see Table 2, calculate the local grain construction operator of inner each pixel of foreground area
(Local Texture LT), sees formula (7), here
The gray-scale value of 8 points around represent pixel piece center pixel reaches,
Operator is the one-dimensional vector of 9 elements, in the formula (10)
For comprising the prospect binary map of shade
Total pixel number,
TIt is the prospect binary map that comprises shade
The average of all pixel gray-scale value partial error averages.The prospect binary map that comprises shade
The zone is at present frame
Pixel
Operator is made as
, comprise the prospect binary map of shade
The background frames of background is not being upgraded in the zone
The relevant position point
Operator is made as
, by relatively these two
If the Euclidean distance of value is less than threshold value
, the present invention gets 0.11, and then this pixel is the direct-shadow image vegetarian refreshments, otherwise is the foreground pixel point, the foreground pixel property value
See formula (11).
HSVColor model meets the visually-perceptible physiological property of people to color very much, can accurately reflect some half-tone informations and color information, for incandescent in the image and extremely dark object, also can reflect corresponding information well.The prospect binary map that therefore, will contain shade
The present frame of corresponding former input
Pixel value is differentiated shade in the hsv color space.The saturation degree of prospect shadow region
The background area of not upgrading with correspondence
Difference less than threshold value
, the present invention gets 0.22; Simultaneously, the colourity of shadow region
The background area of Geng Xining not
Difference is less than threshold value
, the present invention gets 0.08, then is judged as the direct-shadow image vegetarian refreshments, otherwise is the foreground pixel point.
For preventing when background image and foreground object have similar texture information
Method will be ineffective, so the shadow Detection discriminant function needs the simultaneous criterion, sees shown in the formula (11), if comprise the foreground pixel property value of shade
Value is 1, then is shade, otherwise is the foreground target zone, identifies accurate foreground target zone.
Shadow Detection processing block module of the present invention adopts parallel organization, specifically as shown in Figure 2, the shadow Detection frame of dotted line block diagram part representative graph 1, input is the initial background frame
, contain the prospect binary map of shade
, corresponding former input present frame
, output is the foreground pixel property value
8, foreground extraction
Foreground extracting module of the present invention is in order to extract accurate foreground target.With former input present frame
Middle pixel value
Be that 0 zone is left present frame
Original pixel value, rest of pixels is shown as black background, i.e. present frame
The rest of pixels value is made as 0, and the prospect that then will not contain shade accurately extracts, so that succeeding target is followed the tracks of, the carrying out smoothly of pattern-recognition work.
Claims (1)
1. foreground extracting method in the pig house monitor video, in the pig house zone camera system and computer control system are set, there are adjacent symmetric frame difference processing block module, background frames difference processing frame module, motion analysis processing block module, prospect Fusion Module, context update module, shadow Detection module and foreground extracting module in computer control system inside, it is characterized in that comprising following concrete steps:
1) obtains an initial background frame that does not contain prospect by camera system
With the background frames that contains prospect
Sequence also keeps in computing machine;
2) adjacent symmetric frame difference processing block module input background frames
Continuous 3 consecutive frames in the sequence are with present frame
With forward and backward frame
,
Carry out difference respectively and differentiated two binary images carried out AND operation obtaining the adjacent symmetric frame difference
3) background frames difference processing frame module input present frame
With the initial background frame
, be output as the background frames difference of binaryzation
, select the B passage of RGB color space to carry out the background frames difference
4) motion analysis processing block module input background frames difference
With the adjacent symmetric frame difference
Carry out motion state and differentiate, output present frame prospect state encoding
,
It is present frame
Carry out the motion state of adjacent symmetric frame difference computing,
Be present frame
Frame carries out the scene state of background frames calculus of differences; The ratio that accounts for total image area when moving region area foreground moving less than 0.04~0.12 time is slow or static or do not have a prospect; When ratio that the prospect area accounts for total image area background illumination greater than 0.6~0.8 time takes place to become suddenly, do not have prospect in the scene less than 0.03~0.07 the time; The prospect state encoding of prospect proper motion
Be 00, the slow prospect state encoding of motion of prospect
Be 10, do not have the prospect state encoding of prospect
Be 11, the prospect state encoding of background illumination sudden change
Be 02;
5) the prospect Fusion Module is with the background frames difference
With the adjacent symmetric frame difference
Carry out inclusive-OR operation, obtain containing the prospect binary map of shade through filtering, morphological operation and connection marking operation
6) context update module input present frame prospect state encoding
, the initial background frame
, preceding symmetrical frame
With its prospect binary map
, present frame
With its prospect binary map
, output is background frames
When the prospect state encoding
Be to rebuild brand-new background frames at 11 o'clock
When the prospect state encoding
Be 00 or 10 o'clock, with present frame
Remove the prospect binary map
Back remaining area pixel replaces the initial background frame
Respective pixel is by the prospect binary map
The background area of blocking is the initial background frame
The respective pixel value compensates the prospect binary map again
Present frame
Prospect edge pixel and corresponding preceding frame
The average of margin of image element; When the prospect state encoding
It is 02 o'clock, by preceding frame
Remove the prospect binary map
The present frame of back remaining area correspondence
Pixel replaces the initial background frame
Corresponding background pixel, the initial background frame
Middle prospect binary map
The background area of blocking is the initial background frame
The respective pixel value compensates the prospect binary map again
The preceding frame of prospect edge correspondence
Pixel and corresponding present frame
The average of margin of image element;
7) the shadow Detection module will contain the prospect binary map of shade
Corresponding present frame
Pixel value is got
RGBThe value of color model B passage adopts 8 neighborhood territory pixels to make up the local grain zone, calculates the local grain construction operator
And compare two construction operators
The Euclidean distance of value, if less than 0.11, then this pixel is the direct-shadow image vegetarian refreshments, otherwise is the foreground pixel point; The prospect binary map that will contain the direct-shadow image vegetarian refreshments again
Corresponding present frame
Pixel value
Shade is differentiated the saturation degree of prospect shadow region in the hsv color space
The background area of not upgrading with correspondence
Difference less than 0.22 and the colourity of shadow region
The background area of Geng Xining not
Difference then is judged as the direct-shadow image vegetarian refreshments less than 0.08, otherwise is the foreground pixel point;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010105775164A CN102024146B (en) | 2010-12-08 | 2010-12-08 | Method for extracting foreground in piggery monitoring video |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010105775164A CN102024146B (en) | 2010-12-08 | 2010-12-08 | Method for extracting foreground in piggery monitoring video |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102024146A true CN102024146A (en) | 2011-04-20 |
CN102024146B CN102024146B (en) | 2012-11-07 |
Family
ID=43865427
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2010105775164A Expired - Fee Related CN102024146B (en) | 2010-12-08 | 2010-12-08 | Method for extracting foreground in piggery monitoring video |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102024146B (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289819A (en) * | 2011-07-27 | 2011-12-21 | 中国电子科技集团公司第二十八研究所 | Method for detecting infrared motion target in real time for background adaptive estimation |
CN103617632A (en) * | 2013-11-19 | 2014-03-05 | 浙江工业大学 | Moving target detection method with adjacent frame difference method and Gaussian mixture models combined |
CN103957423A (en) * | 2014-05-14 | 2014-07-30 | 杭州古北电子科技有限公司 | Video compression and reconstruction method based on computer vision |
CN104252709A (en) * | 2014-07-14 | 2014-12-31 | 江苏大学 | Multiple-target foreground detection method for look-down group-housed pigs in look-down state under complicated background |
CN105913425A (en) * | 2016-04-08 | 2016-08-31 | 江苏大学 | Self-adaptive oval blocking and wavelet transformation-based multi-pig contour extraction method |
CN107230188A (en) * | 2017-04-19 | 2017-10-03 | 湖北工业大学 | A kind of method of video motion shadow removing |
CN108171740A (en) * | 2018-01-30 | 2018-06-15 | 彭景新 | A kind of method and apparatus based on image analysis vehicle pass-through |
CN108833974A (en) * | 2018-06-29 | 2018-11-16 | 北京奇虎科技有限公司 | Identify the method, apparatus and electronic equipment of the translucent watermark in video |
CN109068099A (en) * | 2018-09-05 | 2018-12-21 | 济南大学 | Virtual electronic fence monitoring method and system based on video monitoring |
CN110235802A (en) * | 2019-07-08 | 2019-09-17 | 北京派克盛宏电子科技有限公司 | A kind of live pig amount of exercise statistical method |
CN111208521A (en) * | 2020-01-14 | 2020-05-29 | 武汉理工大学 | Multi-beam forward-looking sonar underwater obstacle robust detection method |
CN112184773A (en) * | 2020-09-30 | 2021-01-05 | 华中科技大学 | Helmet wearing detection method and system based on deep learning |
CN114943833A (en) * | 2022-03-28 | 2022-08-26 | 台州学院 | Bubble identification image processing method for bubble flow in gas-liquid reactor |
CN115471502A (en) * | 2022-11-01 | 2022-12-13 | 山东土之素生物技术有限公司 | Animal manure rapid detection method based on computer vision |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101763513A (en) * | 2010-02-26 | 2010-06-30 | 成都三泰电子实业股份有限公司 | Foreground extraction method for removing light effect |
CN101854467A (en) * | 2010-05-24 | 2010-10-06 | 北京航空航天大学 | Method for adaptively detecting and eliminating shadow in video segmentation |
-
2010
- 2010-12-08 CN CN2010105775164A patent/CN102024146B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101763513A (en) * | 2010-02-26 | 2010-06-30 | 成都三泰电子实业股份有限公司 | Foreground extraction method for removing light effect |
CN101854467A (en) * | 2010-05-24 | 2010-10-06 | 北京航空航天大学 | Method for adaptively detecting and eliminating shadow in video segmentation |
Non-Patent Citations (2)
Title |
---|
《ACIVS 2009》 20091231 Sonsoles Herrero,et al. "Background Subtraction Techniques: Systematic Evaluation and Comparative Analysis" 33-42 1 第5807卷, * |
《computers and electronics in agriculture》 20081231 Bin Shao, Hongwei Xin "A real-time computer vision assessment and control of thermal comfort for group-housed pigs" 15-21 1 第62卷, 第1期 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289819B (en) * | 2011-07-27 | 2013-05-08 | 中国电子科技集团公司第二十八研究所 | Method for detecting infrared motion target in real time for background adaptive estimation |
CN102289819A (en) * | 2011-07-27 | 2011-12-21 | 中国电子科技集团公司第二十八研究所 | Method for detecting infrared motion target in real time for background adaptive estimation |
CN103617632B (en) * | 2013-11-19 | 2017-06-13 | 浙江工业大学 | A kind of moving target detecting method of combination neighbor frame difference method and mixed Gauss model |
CN103617632A (en) * | 2013-11-19 | 2014-03-05 | 浙江工业大学 | Moving target detection method with adjacent frame difference method and Gaussian mixture models combined |
CN103957423A (en) * | 2014-05-14 | 2014-07-30 | 杭州古北电子科技有限公司 | Video compression and reconstruction method based on computer vision |
CN104252709B (en) * | 2014-07-14 | 2017-02-22 | 江苏大学 | Multiple-target foreground detection method for look-down group-housed pigs in look-down state under complicated background |
CN104252709A (en) * | 2014-07-14 | 2014-12-31 | 江苏大学 | Multiple-target foreground detection method for look-down group-housed pigs in look-down state under complicated background |
CN105913425A (en) * | 2016-04-08 | 2016-08-31 | 江苏大学 | Self-adaptive oval blocking and wavelet transformation-based multi-pig contour extraction method |
CN105913425B (en) * | 2016-04-08 | 2019-02-05 | 江苏大学 | A kind of more pig contour extraction methods based on adaptive oval piecemeal and wavelet transformation |
CN107230188B (en) * | 2017-04-19 | 2019-12-24 | 湖北工业大学 | Method for eliminating video motion shadow |
CN107230188A (en) * | 2017-04-19 | 2017-10-03 | 湖北工业大学 | A kind of method of video motion shadow removing |
CN108171740A (en) * | 2018-01-30 | 2018-06-15 | 彭景新 | A kind of method and apparatus based on image analysis vehicle pass-through |
CN108171740B (en) * | 2018-01-30 | 2020-11-20 | 彭景新 | Method and equipment for analyzing vehicle passing based on image |
CN108833974A (en) * | 2018-06-29 | 2018-11-16 | 北京奇虎科技有限公司 | Identify the method, apparatus and electronic equipment of the translucent watermark in video |
CN108833974B (en) * | 2018-06-29 | 2021-05-18 | 北京奇虎科技有限公司 | Method and device for removing semitransparent watermark in video and electronic equipment |
CN109068099A (en) * | 2018-09-05 | 2018-12-21 | 济南大学 | Virtual electronic fence monitoring method and system based on video monitoring |
CN109068099B (en) * | 2018-09-05 | 2020-12-01 | 济南大学 | Virtual electronic fence monitoring method and system based on video monitoring |
CN110235802A (en) * | 2019-07-08 | 2019-09-17 | 北京派克盛宏电子科技有限公司 | A kind of live pig amount of exercise statistical method |
CN111208521A (en) * | 2020-01-14 | 2020-05-29 | 武汉理工大学 | Multi-beam forward-looking sonar underwater obstacle robust detection method |
CN112184773A (en) * | 2020-09-30 | 2021-01-05 | 华中科技大学 | Helmet wearing detection method and system based on deep learning |
CN114943833A (en) * | 2022-03-28 | 2022-08-26 | 台州学院 | Bubble identification image processing method for bubble flow in gas-liquid reactor |
CN115471502A (en) * | 2022-11-01 | 2022-12-13 | 山东土之素生物技术有限公司 | Animal manure rapid detection method based on computer vision |
Also Published As
Publication number | Publication date |
---|---|
CN102024146B (en) | 2012-11-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102024146B (en) | Method for extracting foreground in piggery monitoring video | |
Bai et al. | Crop segmentation from images by morphology modeling in the CIE L* a* b* color space | |
CN101447082B (en) | Detection method of moving target on a real-time basis | |
CN105740945B (en) | A kind of people counting method based on video analysis | |
CN104992447B (en) | A kind of image automatic testing method of sewage motion microorganism | |
CN109685045B (en) | Moving target video tracking method and system | |
CN102184552A (en) | Moving target detecting method based on differential fusion and image edge information | |
CN105513053B (en) | One kind is used for background modeling method in video analysis | |
CN101299268A (en) | Semantic object dividing method suitable for low depth image | |
CN102842037A (en) | Method for removing vehicle shadow based on multi-feature fusion | |
CN102332167A (en) | Target detection method for vehicles and pedestrians in intelligent traffic monitoring | |
CN102208099A (en) | Illumination-change-resistant crop color image segmentation method | |
CN105046677A (en) | Enhancement processing method and apparatus for traffic video image | |
Keller et al. | Soybean leaf coverage estimation with machine learning and thresholding algorithms for field phenotyping | |
CN101860664A (en) | Moving target detection method robust to illumination variation | |
CN104239886A (en) | Image analysis based lawn and background boundary extraction method | |
Huerta et al. | Exploiting multiple cues in motion segmentation based on background subtraction | |
Naufal et al. | Preprocessed mask RCNN for parking space detection in smart parking systems | |
CN1266656C (en) | Intelligent alarming treatment method of video frequency monitoring system | |
Khan et al. | Comparitive study of tree counting algorithms in dense and sparse vegetative regions | |
Kang et al. | A multiobjective piglet image segmentation method based on an improved noninteractive GrabCut algorithm | |
Koniar et al. | Machine vision application in animal trajectory tracking | |
CN105139358A (en) | Video raindrop removing method and system based on combination of morphology and fuzzy C clustering | |
Xiang et al. | Research on target detection method based on HSV fusion Gaussian mixture model | |
Wu et al. | Nighttime cattle detection based on YOLOv4 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121107 Termination date: 20131208 |